Scribd Logo
Upload

Welcome to Scribd!

  • 20 views

    Emotion-Related Impulsivity and Risky Decisions

    Uploaded by

    Sandra Carolina Valencia

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content

    You might also like

    Emotion-Related Impulsivity and Risky Decisions

    Uploaded by

    Sandra Carolina Valencia
    20 views15 pages

    Original Title

    emotion-related impulsivity and risky decisions

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    20 views15 pages

    Emotion-Related Impulsivity and Risky Decisions

    Uploaded by

    Sandra Carolina Valencia

    Copyright:

    © All Rights Reserved
    Download as PDF or read online from Scribd
    Flag for inappropriate content
    Download as pdf
    of 15
    nica Psychology Review 100 (2023) 102282 ELSEVIER Contents lists available at ScienceDirect 3, Clinical Psychology Review journal homepage: wv elsevier.comilocatelclinpsyehrey Review Emotion-related impulsivity and risky decision-making: A systematic review and meta-regression fe Matthew V. Elliott’, Sheri L. Johnson”, Jennifer G. Pearlstein’, Daniela E. Mufioz Lopez’, Hanna Keren” * Unberiy of Cfo, Bey, Berkey, CA, Und Sas of Arce » al Peay of Mein, Baran Une Safe, ae ARTICLE INFO ABSTRACT ower Emotion-elae impulsivity, the tatlike tendency toward eyretable behavior daring sates of igh emotion ic nation ‘robust predictor of internalizing and externalizing pychopatholgy. Despite substantial evidence that emotion- Inpubviy ‘elated impalsivity fe imporan tranadiognorcally, relatively Iie fs known about its cognitive correlates. This Te a somatng systematic eview end mea egrestan investigates one such candidate, risky decision-making. We analyzed 195 reported in the exginal publications, The mecaregression demonstrated evidence for a smal, positive rla- ‘onehip of emotion-eated impulsivity with behavioral indices of risky decision-making (3 ~ 0.086). Effects generalized across sample age, gender, Postive versus Negative Urgency, and cca versus nonlinieal samples, counting Task, Experimental arousal manipulation was nearly a signifcant moderator, with stess and phat- ‘macological manipulations yielding significant eet sizes. Analyses indicated hat pbliation bis didnot skew" he current ndings. Notwithstanding lintations, the data suggest that risky decslon-maling isa cogaltve domain that relates to emotion elated impulsivity. We coselude with recommendations regarding the specific ‘yper of tasks and arosal inductions that wil bes eaptire emationtlated impulsivity in fatre experimental research 1. Introduction Since the dawn of psychology, scientists and practitioners alike have ‘expressed great concern about impulsivity (Freud, 19475 Guilford and Guilford, 2999), impulsivity generally refers to maladaptive behaviors that oocur without adequate forethought or regard for their conse: ‘quences; however, a singular definition of impulsivity is difficult to achieve given the vast diversity of behaviors ithas been used to describe (oieKlend and Johnson, 2023), In modern psyehologieal science, ‘empirical studies of impulsivity have relied primarily on to forms of ‘assessment (Sharia, Movkon, ane Clark, 207g those that use self report instruments to estimate traitike tendencies, and those tat use cognitive tasks to caprure experimental measures of poor constraint. Although hese measures have all been theorized to tap impulsivity, their coherence is tenuous. Influential seltreport measures, suchas the arratt Impulsiveness Seale and UPPS Impulsive Behavior Seale, have shown that impulsivity can be reliably separated into multiple di- mensions (Pation, Sanford, and Barratt, 1995; Whiteside and Lynam, 2001). Factor analyses of tasks used to test prominent cognitive theories of impulsivity, such as response inhibition, delay of gratification, and risk-taking, havealso not found evidence fora unified impulsivity factor (Reynolds, Ortengren, Richards, and De Wit, 2006; Sharma el, 2014), Furthermore, the selfzeport and laboratory measures of impulsivity have correlated weakly in many eases (Cyers and Cosianpinar, 2011 Shamma et al, 2014), These findings have led co a focus on studying separate dimensions of impulsivity Perhaps the most influential dimensional model of impulsivity isthe UPPS model (Whiteside anc Lynam, 200%). The original model was created using factor analysis and was composed of four dimensions: Urgency, (Lack of) Planning, CLack of) Perseverance, and Sensation Seeking. Confirmatory factor analyses replicated the UPPS model and extended it to include afith dimension, Positive Urgeney (Cie ta, * Corresponding author at University of Calfornla, Berkeley, Deperonent of Psychology, 2121 Shacuck Avenue, Berkeley, CA 94720-1650, United States of mal adres io @erley ss (MN. io), hups//dokorg/10.1016/3.ep-2022.102252 Received 7 March 2022; Received in revised form 12 November 2022; Acepted 23 November 2022 ‘Avaliable online 28 November 2022 (0272°7356/0 2022 The Authors. Published by Elsevier Ltd. This isan open access article under the CC BY lcnse (hp://oeaivecommons 1g /ieenses/>y/4.07. 2007; Whiteside, Lynam, Mille, and Reynolds, 2005). The (Negative) Urgeney and Positive Urgency scales both capture tendencies toward "unconsteained, regrettable speech and behavior during states of high ‘emotion (Carver and Jolson, 2018). (Negative) Urgency items describe ‘contexts involving negative emotions, whereas Positive Urgency items describe contexts involving postive emotions. Although the (Negative) Urgency and Positive Urgency scales are sill often measured separately, they correlate so highly that they form a higher order factor across large scale studies (Carver, Johnson, Joormann, Kim, and Nam, 20115 Cyders snd Smith, 20075 Sperry, Lynam, and Kwapil, 2018), suggesting that they ft together ina dimension that is agnostic to valence. Accordingly, ‘we prefer the term emotio-related impulsivity to denote this traitike tendency t0 respond impulsively to positive and negative emotions. All of the UPPS self xeport scales have strong psychometric properties, Inchuding high test-retest reliability and high correspondence with pparent-eport and interview-based measures (Cylers sn! Sit, 20075 Zapolski and Smith, 20135 Zapolsi, Stirs, Settles, Combs, and Smith, 2010). The strengths ofthe UPPS model have led to its use in thousands ‘of empirical publications over the past two decades ‘Within this vast literature, emotion-related impulsivity has emerged _asaeritial construct in lineal psychology and public health, Emotion. ‘elated impulsivity has been robustly tied to internalizing and exter. rnalizing psychopathology, including substance use problems, aggres- sion, suicidality, borderiine personality disorder, depression, and disordered eating, across 115 studies (Berg, Latzman, Bliwise, and Lil Jen(eld, 2015), with an effect size, r= 34, considerably larger than the effect sizes for other forms of impulsivity measured by the UPPS, which ‘are consistently ator below r =.14. In separate meta ‘related impulsivity has shown moderate effect sizes with bulimia nerv fsa, r = 38, across 50 studies, compared with r < .20 for other UPPS scales (Fischer, Sith, ane Cyers, 2008), and with non-sulcidal self injury, d = 59, again surpassing other UPPS scales, d < .32 across 17, studies (Hamza, Willoughby, and Hetfer, 2015). The effect sizes ‘observed for emotion-related impulsivity are in the range of those ‘observed for other major psychopathology risk factors, such as neuro icism against dimensional measures of depression, r = 0.42 (lou, “Thorsteinsson, and Schutte, 2005). “Multiple other lines of research have validated the clinical import of ‘emotion-elated impulsivity. A growing body of Tongitudinal work has bullt upon cross-sectional research and suggests that emotion-related impulsivity can predict a more severe onset and course of substance use problems, eating disorders, risky sexual behaviors, and self-harm (Gaiser, Bonss, Chamigo, Milch, and Lynam, 2016; Pearson, Combs, Zapoiski, and Smith, 2012; Riley, Combs, Joréan, and Smith, 2015; Riley, Rukavina, and Smith, 2016; Zapolsk, Cyders, and Smith, 2009). “The sale has been shown to predict greater increases in alcohol use after ‘experimental inductions of positive and negative mood states (Cyders, Zapolsk etal, 20105 VanderVeen ct a, 2016). Ina daily diary study, ‘the (Negative) Urgency scale was shown to predict greater increases in seltharm as negative mood states increased (esin, Carter, and Gordon, 2013). Parent and interview-based measures of emotion-related impul- sivity have been shown to have similarly strong associations with psy ‘chopathology (Zapolsk and Smith, 2073), which suggests that effects do not simply reflect biases in self-ratings. Last, there is evidence that ‘emotion-elated impulsivity is heritable, which inreases the likelihood of inter generational effects (Gustavson, Miyake, Hewitt, and Friedman, 20144 Sancher-Roige etal, 2019). Together, this body of work provides ‘additional validation of the scale, and indicates that it has strong, ‘ransdiagnostic predictive validity. ‘The widespread reports of elevated emotion-elated impulsivity in linical populations have prompted experimental studies to determine ‘the basic cognitive processes that correlate with this trait (eg, Cis ‘etal, 20105 Pearlstein, Johnson, Modavi, Peckham, and Carver, 2019), Understanding the cognitive correlates of emotion-related impulsivity ‘would have at least three main benefits. One, it would allow for more precise, controlled, and costelficient research using task-based, cal rycen Review 10 (2028) 102232 functional neuroimaging, which in tum would advance our under- standing ofthe brain systems involved. Two, it would provide a foun- dation to experimentally study the contexts in which emotion-elated impulsivity arises. While the regretable speech and actions that char- acterize emotion-telated impulsivity may be difieult to induce in the laboratory, isolated cognitive correlates could be studied in experiments that manipulate socioemotional contexts like arousal, peer rejection, or hunger and identify areas of special vulnerability. Ultimately, beter knowledge of cognitive correlates, neural correlates, and contextual triggers could set the stage for the development and improvement of psychological interventions for emotion-related impulsivity ‘One might expect that emotion-related impulsivity wotld relate to greater emotional reactivity (Cl, 2005) and indeed, the Negative and Positive Urgency scales have been found to correlate with trait measures ‘of neuroticism (Cyets and Smith, 20085 Sharma e al, 20145 Whiteside ‘inc Lynam, 2001), Many findings, however, do not Fic with the ide that greater emotional reactivity is driving the effects of emotion-elated Impulsivity. For example, those with higher emotion-elated impul- sivity donot show greater psychophysiological or behavioral emotional reactivity in laboratory studies with experimental manipulations of ‘mood (Johnson et al, 2017; Owens, Amlung, Stojek, and MacKillop, 2018; Wise, Phung, Labusehagne, and Stout, 2025). Emotion-eated impulsivity also correlates with psychopathology when controlling measures of emotional reactivity and neuroticism (Cysers etal, 20105 CCyders end Coskunpinar, 2010; Cyders and Smith, 2008), Response inhibition, the ability to override prepotent actions, is one cognitive correlate of emotion-related impulsivity. Although effect sizes of response inhibition and emotion-related impulsivity were larger in clinfal samples (gz, Litlefield, Roslin ad Coffey, 20135 Deker ane Johnson, 2018; Rocha, Beni, Annoni, Vuadens, and Van der Line, 2015), effets have been small in nonelinical samples (Cysers and Cos- kunpinar, 20125 Gay, Rochat, Billie, d’Acremont, and Van der Linden, 2005), and most variance in emotion-elated impulsivity romans un- explained by response inhibition (Joinson, Tharp, Peckliam, Sanches, fané Carver, 2020) Since fast, reflexive processes like emotional reac- tivity and response inhibition do not aecount for most ofthe variance in ‘emmotion-elated impulsivity, more deliberative cognitive processes may alo be involved ‘One such candidate is risky decision-making, which involves mod- lation of goal pursuit based on situational risks or costs. Three func- sional MRI studies have replicated a correlation of emotion-related impulsivity with increased anterior insula activation during decision- making tasks involving risk (Smith etal, 2018} Xiao etal, 20135 Xue, J, Levin, and Bochars, 2010). The consistency of these neuroimaging findings supports the idea that risky decision-making is related to emotion elated impulsivity; however, syntheses of the studies using behavioral measures have been less cleat. In one meta-analysis, Negae tive Urgency was not correlated with measures of delay discounting, whieh tests decision-making for smaller, more proximal rewards versus larger rewards a the est of larger temporal delays (= 0.01, Cyers 9 Coscunpiner, 2011). The meta-analysis included multiple studies employing delay discounting questionnaires but only two studies had included laboratory delay discounting measures with Negative Urgency (Menvearing, 20095 Verdejo-Garci Lozano, Moya, Alcézar, and Pérez aria, 2010), and only one effect size with Positive Urgency was available F = 13, p = 11, Verdejo-Garcia c= al, 2010), Considerable cross study heterogeneity in effect sizes was observed, but the dataset was too small to examine moderation by task format or type. A second ‘meta-analysis resulted ina small correlation ofthe Urgency scale with the lowa Gambling Task (r= 09), and a stronger correlation with delay discounting tasks tht offered actual monetary incentives (F = .24), but ‘ot when such incentives were not present (F = .03) (Sharma et a 2014), Given the limited size ofthe literature atthe time, these findings were drawn from small numbers of effects (k = 2 in some cases). These «wo key meta-analyses did not lead to strong conclusions regarding risky decision-making asa correlate of emotion-elated impulsivity, given the Ihoterogencity they uncovered. Fortunately, dozens of studies with measures of risky decision making and emotion-elated impulsivity have been published since the most recent meta-analysis, opening the door for us to test multiple key moderators. Task type is one such potential moderator, as risky decision ‘making tasks are evidently not monolithic. Research has shown that a ‘common measure of risky decision-making, the Balloon Analogue Rise Task, did not correlate with other measures of decision-making and risk taking, such as the lowa Gambling Task or delay of gratification indices (Gharma et sl, 2014), Conceptual divisions also exist. For example, ‘gambling tasks measure decision-making when consequences are up ‘certain, whereas the consequences ofcholees in delay discounting tasks ‘are certain, Therefor, the types of risk or cost assocated with decisions in these tasks may be important to tease apart Beyond examining task domains, mood state may moderate the relationship between emotion-elated impulsivity and risky decision making. Those with tendencies toward emotion-related impulsivity show more symptomatic behavior after positive and negative mood in ductions (Cyders etal, 20105 Manasse etal, 2018; VanderVeen eta, 2016), and two studies found that response inhibition is impaired after ‘mood inductions for those with emotion-telated impulsivity (Desker ‘and Johnson, 2018; Johnson etal, 2016). Research on mood state as an influence on emorion-related impulsivity and risky decision-making has yielded mixed results n one study, researchers measured risk-taking on the Balloon Analogue Risk Task before and after a positive mood in ‘duction, Positive Urgency predicted a greater increase in vsky behavior after the postive mood induction as compared to pre-mood induction (Cysdess etal, 2010), but these findings were not replicated ina second study (Johns et al, 2026), We alm to synthesize the extant research considering the role of mood in shaping whether those with high ‘emotion-elated impulsivity demonstrate risky decision-making. Given that emotion-related impulsivity relates to positive and negative emotions (which share the experience of arousal), some work has investigated the function of physiological arousal. In one study ‘examining tral-level performance within a response inhibition task, ‘emotion-elated impulsivity moderated the effect of arousal (as ‘measured using pupil dilation) on accuracy. That is, among participants ‘with lower emotion-related impulsivity, accuracy was positively corre Tated increased with higher arousal; whereas among participants with higher emotion-related impulsivity, accuracy was negatively corre lated/ decreased with higher arousal (Pearistcin etal, 2019). Further more, higher Positive Urgency elated to stronger delay discounting ( fe, preference for smaller, more immediate rewards) for imaginary sexual activity (Carrier Emond, Gagnon, Nolet, Cy, and Rouleau, 2018). ‘This fits with the idea that emotion-related impulsivity is particularly likely to lead to unconstrained behavior and decision-making during periods of higher arousal. Accordingly, we systematically review and synthesize studies chat consider arousal manipulations as potential moderators of the link between emotion-relsted impulsivity and risky decision-making, {As the literature on emotion-clated impulsivity has progressed, several other trends have emerged that warrant systematic review inthe ‘current study. One, correlations of emotion-related impulsivity vith response inhibition performance appear more robust in clinical ‘compared to nonelinial student or community samples (Jobson ea, 2016). Two, men tend to engage in more risky decision-making than do ‘women (Byrnes, Miller, and Schafer, 19995 Charness and Gneezy, 2012), and thre, adolescents and young adulis engage in mare risky decision making than do older adults (Steinberg, 2004; Wiloughby, Hefier Good, sind Niagnacea, 2021), Rout, the Positive Urgency and Negative Urgency scales are highly correlated (Cycers and Sih, 2007) and have parallel relationships with many clinical and neuroimaging variables Gohnson, Hliot, and Carver, 2020). Thus, we include sample type (clinical vs. non-cliniead, gender, age, and self-report measure (Positive vs, Negative Urgency) as additional maderators. Taken together, we predict thatthe correlations of emotion-elated impulsivity with risky cal rycen Review 10 (2028) 102232 decision making will be larger in samples that are clinical, have Bigher proportions of young people, and higher proportions of men. We expect no significant difference between efect sizes for Positive Urgency and Negative Urgency. Previous meta-analyses broadly investigating the covariance of ‘impulsivity measures offered the idea that types of rit impulsivity may difer in their cognitive correlates (Cysiers and Coskunpinar, 20115 Shana eal, 201. There now exists the opportunity to extend this work with an eye toward high-priority constructs in clinical psychology. Indeed, the goal of this review and meta-egression is to rigorously investigate the relation between emotion-related impulsivity — a dimension of trait impulsivity of great consequence for clinical psy- chology ~ and rsky decision-making ~ an oft studied, but not yer syn= thesized, cognitive correlate - with special focus on key experimental design and sample moderstors, 2. Method The study protocol and hypotheses for ths systematic review and ieta-regression were pre-registered on the Open Science Framework (OSF) website on May 29, 2020. Along withthe registration (hiips//osi fo/r B27 view only ~3d9a409bc22d672196347412c688626), all data ane. analysis seipts have been posted on OSF (hips//0s"19/n8830/? view-only=6db27a14d8164201bd1ebb9648757700). We note where wwe changed methods after pre-registration. For example, although we hoped to examine curvilinear patterns between emotion-eated impul- sivity and risky decision making, such data was unavailable and could rot be incorporated here. We followed the Journal Article Reporting Standards for Quantitative Meta-Analyses (Sppelbaum et sl, 2018), 21, Search method Yo Identify articles that included an emotlon-related impulsivity measure and a decision-making task, we conducted a literature search fon July 5, 2022. We used five search engines: Psychlnfo, PubMed, Scopus, Web of Science, and psyarXIV. For Psychinfo, our search terms were: (upps OR UPPS-P OR TM(urgency) OR “emotion-rlated impul- sv” OR “three factor impulsiv*” OR “emotion-triggered impulsiv* OR ‘emotion induced impuls*” OR “emotion* impulsiv*") AND (risky OR “probability discounting” OR “intertemporal choice” OR “BART” OR ‘alloon analogue” OR “gambling task” OR “iowa gambling” OR “IGT” OR “delay discount* OR “temporal discount*” OR "choice behavior" OR information sampling” OR “reflection impulsivity” OR “delay of grati- fication” OR “risk-taking” OR “risk taking”). Corollary searches were conducted in the ocher four engines. The searches were limited to arti- cles that were published on or after the year 2001, which is when the UPPS scale and the construct of emotion related impulsivity were pub- lished (Whiteside and Lynam, 2001) Literature searches identified 863 articles, We searched the cited references of these articles and identified 25 additional sties, which were not identified by the literature searches but were potentially, relevant to our review. We also wrote the corresponding authors ofthe articles that were included (see below) to ask for any additional publi- vions and preprints, which yielded six additonal articles and pre- prints. After removing duplicates, the pool contained 621 articles and preprints. 22, Inclusion and exclusion ig. | lusrates the stops by which articles were excluded. To be ‘included, journal articles, manuscripts, and pre-prints were required 10 report empirical data and include both a measure of emotion-elated impulsivity and a behavioral measure of decision-making, Although we considered including self-rated measures of risky decision-making, Wwe see one advantage of behavioral measures being the relevance for fmaging and experimental research, Accordingly, we focus on cal rycen Review 10 (2028) 102232 ‘2022 analogue" OR "gambling tk" taking. ‘enication + Fortis systematic review and meta-epresion we geared a et of ariles tat studied emotion slated impulsivity and sky _Fortapticate sample data 2) task not tigble (ised by arr fier (2) erepot nd tase st in een sample (1) a aaaaaa ‘Mising eect siz and didnot provide in Seah Galanin —. ‘epic on ig. 1. PRISMAstle ow diagram showing the search criteria, selection of studies, inclusion criteria, and exhsion rationale of studies for mete-regresion, behavioral tasks here. The fourth author, DMLL, conducted an intial screening forthe inclusion of an emotion-elated impulsivity measure, ‘ihich led tothe exclusion of 309 articles. MF. J.P, and S.J. conducted independent blind reviews of the remaining 312 articles to determine inclusion in the present mete-analysis. Inerrater reliability for study inclusion was « ~ 0.93, All discrepancies were discussed among the three authors who condkcted blind reviews, and inclusion was deter mined by consensus. 221 articles were excluded for missing a core in ‘lusion criterion, leaving 91 atiles. Anal passthrough the artices led to two exclusions for duplicate sample data, two exclusions for not Including a risky decision-making task (should have been excluded carlier), and one exclusion for having not gathered the self-report and task data within the same sample. Of the 86 studies that met fll eiteria for inclusion, 30 studies reported a total oF 80 relevant eect sizes with the corresponding sample sizes and moderator variables in the main ‘ext, the supplement, or in an online database. 56 studies did not report any type of statistical test ofthe relationship between emotion-related {impulsivity and decision-making task performance. For these studies, wwe emailed corresponding authors to request the data needed for in- clusion. The authors of 2 studies replied withthe requested effect sizes for av datasets From these 21 studies we added 108 effects, which are reported here for the is time, The authors of $5 studies didnot reply to to attempts at correspondence ot responded that they did not have Information to share, therefore those studies could not be included. The final set of n= 51 studies yielded k = 195 effet sizes. 2.3, Data extraction 23.1. Bifect sees ‘MLE, S.J and J. extracted allelevant effec sizes forthe strength of ‘the relationship between emotion-elated impulsivity and risky ‘decision-making. In one exception o including multiple effect sizes from ‘given study, we chose to use only the frst chronological effect reported ina three-wave longitudinal study (Soothe), 2017). Most eect sizes were bivariate corzelation coefiients, however, patil corelation co ‘efficients from one study were also included (20 et 3, 2008). Where raw datasets were provided, we calculated Pearson's for each relevant effect. We converted correlation coefficients to Fisher's zto contol for undesirable statistical properties in corelaton coefficient distebutions. 2.3.2. Variable operaionalizaion and coding ‘We coded study features inthe following steps, Fist, MA, J.P, and ‘Sd.each coded a third ofthe included studies independently. Second, M. . and JP. reviewed the coded studies, made coding adjustments as needed, and recorded the rationale forthe adjustments. Third, SJ. and MEE checked all studies, including reasons for any coding adjustment. Interrater reliability of coding across all variables was very high, « = (0.99, Rare disagreements, such asthe coding ofa sample as clinical vs. non-clinical, were resolved through discussion and consensus. To assess descriptive features, we extracted the sample size and ‘coded publication status. As proxies for sample diversity, we coded the ‘country where the study took place and racial identity. The race variable was operationalized asthe percentage of study participants that ide ‘ified as white; we chose tis index due to heterogeneity in how race was reported. Next, we coded the preegistered potential moderators. For the _gender variable, che proportion of men and women was operationalized as the percentage of women in the sample. The age variable was oper- _tionalized asthe mean age of participants in the sample. For the sample type variable, samples were coded as (1) clinical or (0) non-clinical (which included community and undergraduate samples). We also ‘coded the self-report measure used in exch effect size ~ Positive Urgency (PU) or Negative Urgency (NU). To operationalize task-type inthe mete "regression, we assigned each risky decision-making task an abbreviated code (e IGT for lowa Gambling Task; !sble 1). In post-hoc analyses, we grouped effects from all "gambling tasks." which were defined as ‘tasks where participants wagered something of value ona trial with an uncertain outcome. The following tasks were re-coded as “Gambling” in this secondary analysis: Cambridge Gambling Task (CGT), Columbi Card Task (CCT), Cups Task, Game of Dice Task (GDT), Holt-Lawry Risk ‘Task (Hol), Iowa Gambling Task (IGT), Risky Gains Task (RGT), and Verbruggen Gambling Task (VGT), ‘The inclusion and type of arousal or mood induction were coded sing the following categories: (1) no manipulation of arousal or mood; (2) imaginary sexual incentive; (3) peer supervision; (4) pharmacolo {cal manipulation of arousal (., yohimbine); (5) stress induction; (6) positive mood affect induction. Separately, we coded whether monetary incentives were provided based on performance. Tasks in which par Bipants only earned points oF fictitious money were coded as not providing monetary incentives Some studies included multiple tasks and/or multiple indices of risky ‘decision-making per task. For these studies, we analyzed data for each respective index. One exception was the Balloon Analogue Risk Task. ‘Because balloons are set to explode at random, the number of exploded balloons is less directly reflective of participant behavior than the number of pumps on unexploded balloons (isjuer, Aklin, 2volensky, ‘sod Pedlla, 2003). Therefore, we prioritized the number of pumps on "unexploded balloons and ineluded the number of exploded balloons only when number of pumps was not availabe (k= 2). For each index of sky decision-making where high scores represented low risk-taking, we reversed effect sizes so that high seores would consistently indicate high risk-taking (¢g., ~0.2 changed to 0.2). cal rycen Review 10 (2028) 102232 Tablet Description of tasks aon Desinon sth ed ah wot hey "hey athe one se Hoh an fe "nkrpt Risk Tsk AR falloe o reve mere ey, ienpode. contig ogee, 1999 The patpne gues wether ‘ambing Tse 2 yelow token biden ned con fe blue bx outa es psie tenes ‘coumbia Card Fig, Maca, Parga sequal dele Task (GCT) Whey anc Weber how many fre dows cad 0 3ro tum ower Wincor ado ter wns we tuning ove Inet seas oir tea and rerminae he ‘ups task Levin and Hart, 2063 "The pan ehaees Benson aeration that nvever tw ove cape at he parteant mst chose "ase BOP) bento an este al Driving since frown tal, 2016 The part ives a Hiprecy reson condition ‘rand a, 2005 The panna peti the on tear of ie ol by ecg between ots Wil "rk en apie ow protailiy read dow oie, igh peobabiliy normation cr Robin tsehe, The pripat chooses ‘Sampling Tak ond Soskion, 2006 aan ore re foraen ost, efoe mati a deco, town Gambllng ssh, Tonel and The parla hover ead “ek OG) Dara, 2000 trom four docks which eer reward o penalize the hie Teaing tat some des te ese tan te oer CObsacecoome owen Sowa, Aye ek tak bee Rds ‘one ope greta and tran rg eaabte ala, 2018 fours a inry ik eave sky Gar Task Kosch, Sines oer, Fagan, snd Pela 2012 alate nein ower Single ey Dough, Mit The print obtains ware Parti SHOP) esos reponse Partie (TCP) ‘cuenpnid bya longer delay shorter dey Vero Vesboggen, Chambers, Papal heal ‘Canblng Task Laveen ed Mela, ard hae gated of won, an? ambl on are ead thats ‘We considered multiple standard guidelines for assessing study quality (Valentine, 20195 Viswanathan and Berkman, 2012), Because these are often developed for medical and intervention science, frequently coded indices, such as interrater reliability, blinding, random assigament, and inclusion of comparable control groups or placebo were not relevant to the current analyses, In many ways, our studies were already constrained by our requirement to focus on behavioral indies of risky decision-making. Nonetheless, we identified and coded one key indicator of data quality: the number of trials in each risky decision-making task, which can influence task reliability. Beyond data quality, studies differed in two facets of their reporting quality ‘whether they reported the racial identities oftheir participants, which may indicate attention to sample repeesentatveness, and in those studies that included an arousal or mood induction, whether they re ported data about the succes ofthe manipulation. 2.4, Planned analyses Alpha was set at 0.05 and analyses were conducted using R. 2.4.1. Meta-regression - pooled effect and moderator The pooled effect size between emotion-elated impulsivity and laboratory indices of risky decision-making was estimated using mete regression. An advantage of mets-regression isthe ability to examine moderating effects concurrently in a single model (Psstsjovsky and “ipton, 2022). Much like multiple regression in single sample studies, this improves precision in estimating the effect of each moderator on the ‘outcome variable. Traditional random effec meta-regression assumes independence between effect sizes. Therefore ifthe withn-study covariance structure is unknown, as is offen the case in metaegression, multiple, non independent effect sizes from a single study must be pooled. This ca ses information loss and reduces the ability ofthe researcher to examine sources of heterogeneity. Robust vatiance estimation (RVE) is @ mete ‘analytic technique that is robust to dependent effects in meta-analysis and provides valid estimates and standard errors even when the un derlying dependence structure is unknown (Pp0, 20 Viton, 2015), Because we did not have access to the within-tudy covariance struc tures from studies that reported multiple effects, we shifted from our pre-registered plan of using structural equation meta-analysis to RVE meta-regression For our core analyses, we used the robumeta package to build two correlated effects RVE meta-regression models with assumed within sample correlations (eho) set to 0,6 (Pisher, Tipton, and Zhipeng. 2017). To testrobusines of he findings, we also bail equivalent models ‘with cho equal to 0.4 and 0.8. The frst meta-regresson was an intercept-only model to estimate the pooled effect size. Studies varied in the numberof relevant effect sizes ‘that they reported, and five studies reported more than ten effects and ‘accounted for 80 out of 195 total effect sizes (419%), To determine whether the findings of the original model were skewed by these five studies being overrepresented, we tested the intereept-nly mete regression with these studies removed (not pre-registered). The second meta-regression model included moderator variables of interest — age, gender, task type, arousal manipulation, monetary incentive, sample ype, and impulsivity measure type. We followed two ‘uldelines to avoid interpresing underpowered moderator. Fist, any ‘moderator variables that were missing for the majority of effect sizes ‘were dropped from the meta-rgression. Second, per recommendations by TannerSinith, Tipton, ané Polenin (2016), we did not interpret moderators with fewer than four smallsample corrected degrees of freedom, For categorical moderetor variables, any levels that did not Ihave four or mote smallsample corrected degrees of freedom were collapsed to conserve data Recent advances in the flexibility of RVE meta-regression have allowed for models that simultaneously account for correlated and hi cerarchical dependence structures ina dataset (Psstejovsky and Tipton, 2022), We built a correlated and hierarchical effects (CHE) model with [RVE standard errors using the metfor (Viechtbascr, 2010) and club Sandwich (Pastejvsky, 2020) packages and the methods described by Pusiejovsky and Tiptos (2022) to test the robustnes of our moderation cal rycen Review 10 (2028) 102232 findings inthe case that unforeseen hierarchical dependence structures existed in thi dataset. For each ofthe models described, the relevant effect sizes. the 2 transformed correlation coefficients) were included and weighted pro- portionally to their sample variance, which was estimated as an adjusted {inverse ofthe sample size: V = qb (annerSiit ta, 2026). Standard errors were then estimated ax the square root ofthe variances: YAR Gorenstein, Hedges, igs, and Rohsten, 2009) "We catinate heteropensty of effing tw commen meses? and P. x? is an estimate of the between-study variance relative to the withinstudy variances. desribes the percentage of between-stdy aimee that in due to Detrogenciy compared bo stapling roe Gistins ane thomson, 2002) Substantial heterogeneity (F > 50%) may be indcaive of meaningful subgroups in the st of eet sizes Giigins ra, 2019) 24.2. Non hypothesized sample characteristics In adation to age and gender, which were pre-egisered moderated variables included in the meta-regression, we tested Une generalizability of effects in elation to three sample characteristics that were not pre registered — percentage of the sample identifying as white (ie. Percent White), year of publication, and country of research. These were rot hypothesized to moderate the aggregate effect size, and we tested each separately using the corzelated effets RVE meta-tegression tech nique deseribed above. 24.3. Suudy quality ‘We also tested the stability of effects in relation to three data and porting quality indicators: 1) umber of tras in the task, 2) whether ‘the racial identities ofthe participants in the sample were reported, and 13) whether studies that used arousal manipulations checked their effectiveness. We tested each of these separately using the correlated effects VE metaregression technique described above (not pre- registered), 24.4, Assesment of publication bias ‘We investigated potential publication bias in four ways. First, we used a funnel plot to visually inspect the data and examine for skewness in the bivariate distribution of effect size and sample size. Funnel plots reflecting publication bias are asymmetrical and have disproportion- ately few nonsignificant effects with small sample sizes (Steme on ‘arbord, 2004). Second, we eondueted a trim and Ml analysis (Duval and Tweedie, 2000), wsing the meta package (Baldzri, Ricker, and Schwarzer, 2018), We elustered th effects by study and chose a random effets model to estimate the number of unpublished null effet sizes, siven the distribution of effects included in this meta-regresson. Third, in addition tothe pre-registered checks, we conducted a P-Curve anal ysis (hs://p-curve.com) to examine the distribution of significant p- values and test for evidence of a true aggregate effect (Simonsohn, Nelson, and Simmons, 2014; Simonsohn, Simmons, and Nelson, 2015), Inthe case ofa true effect, the p-curve is right skewed with an increased, incidence of low significant p-values (p < .01) compared to high sig- nificant p-values (04 < p< 0.05). Fourth, we conducted a sensitivity analysis using the PublicationBias package to determine the severity of publication bias that would have to be present to fully attenuate the aggregate effect (Mathur and VanderWecle, 2020). Severity of publi- cation bias (7) was defined as the number of times more likely an affirmative study was to be published than a non-affirmative study and tested ata recommended range of 1:0 200 3. Results 1, Study characterises Appendix A includes sample characteristics, moderator variables, ‘and study quality indicator forthe n ~ 51 ineluded studies, yielding k 195 effect sizes that reported dats from 14,957 participants (age M = 32,94, SD ~ 10,62), 90 effect sizes were extracted from an article or supplement, and we generated 105 novel effect sizes from studies that had the necessary data but had not reported on emotion-telated impulsivity and risky decision-making. Of the 195 effect sizes Included, only five were covered in previous meta-analyses (Ces an Coslanpinar, 2011; Sharma ers, 201-9.'The numberof included effet sizes per study ranged from a minimum of one to a maximum of twenty (median = 2). Five studies reported ten or more effect sizes; however, most effect sizes eame from studies that reported four or fewer (AP: pendix B.1). Figure 2 illustrates the distributions of the continuous sample characteristic variables ~ Age, Percent Femsle, Percent White, and Year of Publication. Table 2 captures the numbers of studies (n) and effect sizes (k) of the categorical moderator variables - Sample Type, ‘Arousal Manipulation, impulsivity Measure, and Task Type. Four tasks hha sufficient data to be included as independent eategories ofthe Task ‘Type variable — Delay Discounting Task (n = 23, k = 83), Balloon ‘Analogue Risk Task (n = 13, k = 27), lowa Gambling Task (n= 9, k 24), and Information Sampling Task (n= 4, k= 14).To conserve data for the tasks that were used in fewer than four studies, the remaining twelve tasks were included with the task type variable re-coded as “other” (n= Mean Age 1 ¢ Percent Female Gender cal rycen Review 10 (2028) 102232 13, k= 47). Because several variants of gambling tasks were included in the “other” category, we conducted post-hoc analyses ofa consolidated category of gambling tasks, No individual type of arousal manipulation achieved four small-sample corrected degrees of feedom, so we reduced, ‘ur sx eategoris to examine whether there vis an arousal manipula: tion (1) oF not (0) 3.2, Intercept onty meta regression ig. 9 lusates the distibution of effect sizes across all studi. Using an intercept only meta repression, we found asmalaszed eect of sky decision-making task performance on emotion-eated impulsivity (6 = 0.086, ¢=5.72, p< 0000). Effect size estimates (Fisher's 2) fort Individual studies ranged from ~0.48 to 0.60. "The proportion of var- ability in effet sizes from heterogeneity relative to variability from tamplingeror was moderate thigh, = 60.2, which provided eve dence for the import of considering moderator eflets (sss et 3, 2019). The between-study variance In effect ies was small elative to ‘he within. study variances (= 0.0062), which justified the inclusion of @ CHE model to test for robusinest in the event ofa hierarchical Covariance structure When the five sties that reported more thn ten effet snes were removed, the intereeptony model yielded a parallel salad significant effect (8 = 0.088, ¢= 5.53, p< 0000) 3.3. Moderated meta regression “Table 3 includes model ousputs from the full RVE meta-regression Percent White ication gE 3 V Year Fig. 2 Violin pots of A) the mean age () the percentage denying as white and (C) the percentage identtyng as female (D) the year of publication serosal samples incided in the meta regression ‘Table Frequency of categoria! ma Sane ope Comey 2 to 060 Ars! mann Nepuive wren ” ne fare ‘Nowe BART ~ Balloon Analogue Rsk Task, OCT = Columba Cad Task, GGT = Cambridge Gambling Task, DDT = Delay Discounting Task, ER! — Emotion related impulsivity, GD? ~ Game of Dice Tas, IGT ~ lows Gambling Takk IST = Information Sampling Tas, K = number of effect sles, ~ number of studies, RCT ~Rishy Gains Tash, SKIP ~ Single Key impulsivity Parada, TCP Two Choe Impulivity Paradigm, VGT ~ Verbruggen Gambling Tas With modertors included. The strength of the correlation of risky decision-making with emotion-elated impulsivity was moderated significantly by Task Type. In aggregate, scores from the Balloon “Analogue Risk Task (B = 0.050, ¢ = 1.71, p = 0.114) and Information Sampling Task (8 = 0.016, t= 0.29, p = 0.783) (Appendices B.2-3), did ‘not correlate significantly with emotion-elated impulsivity scores. Scores from the Iowa Gambling Task (8 ~ 0.197, ¢ = 426, p ~ 0.011) (ig. , Delay Discounting Task, (6 - 0.108, ¢ — 5.78, p < 0.001) (ig. 5), and tasks coded as “other” (8 = 0.145, t = 2.98, p = 0.017) significantly correlated with emotion-telated impulsivity scores, The Arousal Manipulation moderator showed a nonsignificant trend (@ = 0.109, = 238, p = 0.051). As shown in Table 3, Monetary Incentive, Sample Type (clinical vs, non-clinical, Measure Type (Negative vs. Positive Urgency seale), Gender, and Age were not ig nificant moderators of effec sizes In the full model, the proportion of Variability in effect sizes from heterogeneity relative to variability from sampling error reduced to a moderate level, P = 47.89. The between study vatiance was & = 0.0047. Results were parallel when testing higher and lower values of cho (Appendix C.1-2). The Correleted and Hierarchical Effects (CH) model ‘also had vey similar results: however, this version ofthe model yielded ‘2 nonsignificant cumulative effect size forthe asks coded as “Other” (S 0.124, 959% CI [-0.003, 0.251]) and a weaker effect of Arousal cal rycen Review 10 (2028) 102232 Manipulation (8 = 0.080, 95% CI [ 0.064, 0.223) (Appendix C3). ‘When using the re-coded task type variable, the pooled effect was sig nica fr the gambling tasks (8 ~ 0.158, ¢ ~ 3.84, p ~ 0.003). In this instantiation of the model, the tasks coded as “Other” (8 = 0-180, 2.63, p = 0.050) bordered a significant effect, and the Aroussl Manip ulation moderator achieved significance (= 0.101, = 2.37, p = 0.049) otherwise, the indings remained the same (Appendix C.4). 34, Glaser look at arousal and mood manipulations Given the relatively small number of studies that experimentally ‘maniptilated mood or arousal, none of the unique maniptlation types had sufficient power to be examined independently, Despite the limited statistical power, we report effect sizes for different types of mood/ arousal indutions separately (she 2. Studies that induced arousal via stress, imaginal sx, or pharmacology (ic. yohimbine) reported signif= ‘cant effects within the conditions involving arousal manipulations 3.5, Publication bias “To assess publication bias, we frst visually inspected a funnel plot of all included effect sizes (xe 6). Our distebution appeared approxi- mately symmetrical, with small and large sample studies being repre- sented by positive and negative, nonsignificant and significant effec, However, our trim and fil analysis added 30 effect sizes. This indicated a small asymmetry inthe funnel plot and justified further evaluation. Our pecurve analysis showed # rightskewed distribution of significant p= values that indicated an absence of systematic “p-hacking” and publi- cation bias (Appendix B.4), such that a true mul effect inthis literature was highly unlikely (Binomial, p 0007; Continuous (Stoutfer), 2 =6.48, p< 0001). Furthermore, the evidential value was determined to ‘ot be inadequate or below the 339 power threshold (2 = 1.88, p (one sided) = .971), The power of the tests included in the p-curve was estimated to be 53% (90% CI: (368%, 68%]. Fourth, our sensitivity analysis concluded that given the 41. affirmative and 154 non- affiemative effect sizes in the dataset, even the most extreme level of publication bias tested (y = 200) would not fully attenuate the signifi- cant pooled effet (Appendix B.S). 36, Bcaminaton of non hypothesized sample characterises and study ‘nualiy There was no evidence that non-hypothesized sample characteristics ore related to effec sizes (Table 5). The racial identities of sampled participants were only reported for 44.1% of effect sizes. Across those studies that did report on racial idensties, there was no effect of non- White race representation on effect size. The year of publication for the included studies ranged from 2005 to 2021, and it was not signfi- cantly related tothe effect size. 66.236 of effect sizes came from studies chat were conducted outside ofthe United States. There was no signi= fant difference in effect sizes from the United States vs. other counties, "The number of rials was not significantly correlated with effect size across the ful set of studies (Vable 8). Regarding porting quality, the percentage of participants identifying as white vas reported for 86 of 195 effect sizes. Effects coming fom studies that reported on racial fdentty were on average modestly smaller than effets that came from. studies that did not report on race (8 = ~0.073, = ~2.53, p~0.016). OF 7 15 effect sizes from studies that included an arousal manipulation, 9 came from studies that reported information about whether the arousal ‘manipulation succeeded; all reports confirmed suecess. We did nat have enough effect sizes from these studies to conduct a statistical test. 4, Diseussion ‘Our systematic review of published and unpublished research on ‘emmotion-elated impulsivity and risky decision-making uncovered a rich cal rycen Review 10 (2028) 102232 “10 0s ° os 10 Effect Size (Fisher's 2) 23, Forest pot for al inched eects of risky decision-making and emotion-elated impulsivity, Point eximates (squares) with 95% confidence intervals (horizontal lines) are sorted from strongest postive (top) to srongest negative (botom). Positive effect sies are those that found high risky decision-making ‘comelating with high emotlon-elataé impulvty. Solid vertcl line represents the pooled effec, extimated by Intercepconly mea-egreson. Dashed ver lines represent the 95% confidence interval forthe pooled ff 7 ‘Table Covrlated effects, RVE moderated met-opresson predicting elect size (isher’s 2), tho = 6. Moderator ee 2 SE n = z 35h bow Ser Task pe BART ‘050 bam Tn one “aor ona Task pe DOT 10s oo as aon ne? cory “ose pe IGT o1s7 Dose a oo oor om Task pe (ST 016 base tn 7m “xs ote raskepe hee ones bow a8 oats ome 0336 Percent oa “002 bo net ose “Soe 000 ae 02s pois 1422 ora 0038 cinta “0% rr) ‘ie O38 Maney a sat sts 10m Yea 017 bois 1700 oie conn? Dow 00 mes tae epee rc 0073 a9 ms ais? “21 ‘Now* = ested separately, BARE = Balloon Analogue Ws Task, Ci. = Confidence interval, DDT = Delay Discounting Task, df= degrees uf eal, HR related impulsivity, {GT ~ lw Gambling Task, RVE = Robust Variance Esti literature with diversity in sample demographics, geographic location, ‘experimental tsk type, an clinical status. Indeed, beyond synthesizing the published correlations of risky decision-making with emotin- related impulsivity, thie study was generative in nature. Over 20 col leagues re-analyzed or shared their raw data, allowing us to add 105 effect sizes to this review that were not previously published, an add ‘on that greatly improved our ability o consider risky decision-making and emotion-related impulsivity effects. ‘We used a RVE mete-regression approach which provided two core fon, YIN Binary variable (he, “yes/0) strengths over other meta-analytic approaches. We were able to conserve data from studies that reported more than one relevant effet size, and we were able to examine the relative effets of maltiple pre- registered moderator variables concurrently. Across $1 studies and 195 effect sizes of nearly 15,000 participants, we found a small effect between emotion-related impulsivity and increased risk-taking behavior fn lsboratory decision-making tasks. This effect, though small, was robust. It oes not appear thatthe results were confounded by publi- cation bias. The effect has held steady over the past two decades and cal rycen Review 10 (2028) 102232 Effect Size (Fisher’s z) ig. 4, Fores plot for all ncuded effec from the Lowa Gambling Task. Plnt estimates (squares) with 95% confdence intervals (horizontal nes) ae sorted frm, strongest positive (tp) to strongest negative (bottom). Positive elect size ae those that found high ky decision-making correlating with high emotion elated. impulsivity. Solid vertieal line epresents the pooled effect, estimated by moderated meta-egresson, Dashed vertical lies represent the 95% coafidenceinterral for the poled effect ‘generalized to studies conducted outside the United States and across ‘gender and age. Regarding study quality indicators, whether researchers Teported participants’ racial identities did relate to effect size, whereas the numberof teal in the task didnot. AAs hypothesized, the metacorrelation between sky decision making tasks and emotion-related impulsivity was comparable for Pos itive Urgency and Negative Urgency. This is consistent with recent ‘theory Coinson eal, 2020) and empirical work showing that Positive Urgency and Negative Urgency are themselves highly correlated and load onto a supraordinate factor (Carver etal, 2011; Cyéers and Smith 2007). “More surprisingly, there was no evidence for moderation by mone: tary incentives, and the evidence for moderation by arousal manipula ‘tion was borderline. As only 15 sucies included arousal manipulations, we aggregated the forms of arousal manipulations in the mets regression. Closer examination of the unique effects indicated that Positive mood induction (with perhaps the exception of imaginal sexual rewards) did not strengthen the link between emotion related impul- sivity and rsky decision-making. Tentatively, this may reflect that many laboratory postive emotion induetions have small effects on arousal most participants will not experience large amounts of excitement ‘while completing cognitive tasks in a laboratory setting, Although ‘power was limited, effet sizes from the single studies of stress induc ‘ion, imaginal sexual rewards, and yohimbine, a pharmacological arousal induction, were significant, These three techniques represent arousal manipulations involving negative valence, postive valence, and without regard to valence respectively, This aligns withthe idea that arousal, rather than valence is the component of emotion that intersects with loss of sele-contel inthis type of impulsivity (Jolinso» el, 2020), Future work on understanding arousal moderation is of particular importance inthis literature. Contrary o our hypothesis, there was no difference in the aggregate cffoct sizes of studies that used clinial vs. non-clinical samples. This mull, effect contrasts a meta-analysis that observed a higher correlation of response inhibition performance with emotion-elated impulsivity in clinical samples (* = 0.34) then community (r= 0.14) and student (0.11) samples (olson el, 2016), and with evidence that response Inhibition deficits show a curvilinear pattern, in which they are more closely correlated with emotion-telated impulsivity within the higher ranges of emotion-elated impulsivity (Dekker and Johnson, 20185 Johnson et al, 2016). Tentatively, it may be that ncuropsychological correlates are only explanatory atthe more severe range of ths form of impulsivity, whereas more deliberative decision-making indices could be correlating across a fuller range of emotion-elated impulsivity. This {dea could be tested more systematically in future research. ‘Out of the four most common risky decision-making tasks, perfor- ‘mance onthe Balloon Analogue Risk Task and the Information Sampling, ‘Task didnot significantly correlate with emotion-related impulsivity. On the other hand, the Iowa Gambling Task, the Delay Discounting Task, and an aggregate of other less comumon taskiypes were each correlated ou i OS cal rycen Review 10 (2028) 102232 05 Effect Size (Fisher's 2) 5. Forest plot forall included eet frm the Delay Discounting Tatk Pot etinstes(aquare) with 95% conBence intervals (horzoatal lines) ae sorted fom ssrongest positive (top) to strongest negative (boom). Positive eles sizes are those that found high risky decision-making correlating with high emoton-eated. impulsivity. Solid vertical line epresents the pooled effec, estimated by moderated metz-egresson. Dashed vertical lines represent the 95% coafidence interval {or the poled effet With emotion-elated impulsivity. Caution is warranted about inter preting the aggregated effect size for this heterogenous set of “other” tasks; however, the data suggest that tasks beyond the four most com: ‘mon are worthy of consideration for future researchers, ‘Our ability to compare effects by task type shines new light on the ‘cognitive correlates of emotion-reated impulsivity. Our findings in ‘cated that the preference fr gaining rewards more quickly, as measured ‘on delay discounting tasks, is related to emotion-elated impulsivity. Delay discounting tasks provide the purest metres of decision-making in ‘this Iterature because they yield isolated time discounting parameters, ‘derived from se of certain and unambiguous deesions. The other asks blend decision-making with varying degrees of ambiguity and uncer tainty. Among these are the “gambling tasks.” The CGT, Cups Task, GDT, Holt, RGT, and VGT all Involve decision-making under conditions of ‘uncertainty, but not ambiguity, That is, the probability contingencies of ‘hoies are known, but the outcome of any single tral is unknown, The IGT and CCT are unique among the "gambling tasks" because the probability contingencies ofthe posible choices are unknown and must be learned implicitly through feedback. Therefore, the IGT and CCT are both uncertain and ambiguous. When we included all "gambling tas’ a a single task type, the isolated effect we found for IGT with emotion. related impulsivity remained. From these studies we have evidence that the link of emotion-elated impulsivity with risky decision-making impulsivity generalizes across laboratory tasks with varying levels of “uncertainty and ambiguity, This generality of the effect across tasks with varying levels of ambiguity and uncertainty may signal thatthe real- World manifestations of riskier decision-making occur in a wide array ‘of contexts for those with mote severe emotion-related impulsivity. Unlike the DDT or "gambling tasks,” in which participants have an explicit or implicit understanding of the relative risk and reward for making riskier choees, the BART tests ssk-taking ina purely stochaste system. This randomness, and consequent lack of available strategy for all participants, may be the ingredient that sets the BAR apart from other risky decision-making tasks that do appear to capture some ofthe variance in emotion-rlated impulsivity. Our findings appear consistent with findings of one meta-analysis (Sharm et al, 2014) in whleb a facior analysis showed that the BART did not load onto the same Impulsive Decision-Making” factor as the IGT and DDT. Although we found evidence that the BART provides a poor laboratory measure of ‘emotion related impulsivity, there was an absence of BART studies using arousal induction methods that appear most promising (©, ste, yohimbine). The Information Sampling Task provides an interesting parallel here. Like the BART, the pooled effect for che IST nificant, yet in the only study with a pharmacologi Standard Error cal rycen Review 10 (2028) 102232 Effect Size (Fisher's z) 6, Funnel pot for studies examining effects of risky decision-making and emotion slated impulsivity. Vera dased line represents the pooled effect size ‘Doted diagonal lines (atror) represent the 95% confidence terval. Dashed diagonal lines (exterior sepresent the 9% confidence Interval manipulation, yohimbine administration significantly moderated the relationship between emotion-related impulsivity and risky decision making on the IST (Herman, Critchley, ané Duka, 2019). More work is needed to test whether risky decision-making. task performance, including the BART, correlates with emotion-related impulsivity in the ‘context of strong arousal manipulations ‘Taken together, we observed a highly significant correlation of, ‘emotion-elated impulsivity with risky decision-making, which was small but replicated across many types of laboratory tasks. The small ‘aggregate effect size across tasks mirrors a broader issue of oor ‘concordance between trait measures an laboratory experiments ofthe same construct (Dang, King, 2nd Invicht, 2020), One possible expla nation is that effect sizes in this literature may be attenuated by the reliability of risky decision-making tasks, as some gambling tas (Le, IGT and GDT) have shown low test-retest reliability (Suciow and ‘Barnhart, 2018). This would not appear tobe a good explanation forthe low effect sizes across tals, though, as the DDT and the BART have shown acceptable test-retest reliability (Anokinn, Goloshevisn, ane ‘liga, 2015; Weafee, Baggoxt, and de Wit, 2013), Beyond reiabity, it is possible that tasks and self-ratings eaprure fundamentally different processes and contexts. For one, there is the necessary methodological disconnect between trait measures which capture “on-average” len dencies across time and “snapshots” of behavioral performance in the laboratory. Together, these factors might lead to an underestimate inthe true relationship between the constructs of risky decision-making and ‘emotion-elated impulsivity. 4.1. Comparison with previous reviews ‘This meta-tegression builds on nearly two decades of research onthe laboratory correlates of emotion elated impulsivity, inluding relevant meta-analyses (Cyders and Coskunpinar, 2011; Sharma et al, 2014), ‘The Gye avd Coskunpinar (2017) review was restrited to examining, ‘wo studies of laboratory delay discounting, and no other risky decision- ‘making tasks were included. Shara l. (201) expanded this work to {include additional rsky decision-making tasks, and (o differentiate ef fect sizes by whether rewards were real or hypothetical. The current work extends the findings of these two pioneering meta-analyses by adding 190 new effect sizes, including additional task types, key experimental and sample moderator variables, and greater representa tion of the Positive Urgency scale, modeled together using recent RVE smeta-regresson techniques. 4.2, Limitations Despite the strengths, there are several limitations to the present work. Fist, the generalizability of the findings with regard to racial Jdentty and lifespan development must be interpreted with caution. Our ‘operationalization of race 2s “white” oF “non-white” was chosen 10 conserve data inthe face of different reporting details, but i provided no granularity t@ examine generalizability across racial identities Furthermore, most studies did not report race, and those tha did report ace — which we view as one indicator of the quality of reporting - hed. smaller effet sizes on average. Also, most samples included only adult participants, so there is insufficient evidence to know if Bindings ‘generalize across the lifespan, Second, we lacked statistical power to ‘compate some levels of some categorical moderator variables that may’ ‘exert distinc influences. Third, some moderator variables inthis dataset fare not distributed evenly across the categories of other moderator variables eg, strong arousal manipulations were more common in DDT than BART studies). This is largely accounted for by the structure of the rmeta-regression; however, it remains formally undifferentiated. Fourth, we did not conduct a forward citation search, and itis possible that we missed one or more relevant findings 43. Pure directions Drawing on our findings, we conclude with four discussion points to help guide future work in thie area. First, our findings suggest that ef fects of risky decision-making are present, but modest. Alongside response inhibition, risky decision-making is a cognitive domain that correlates with emotion-telated impulsivity, indicating that emotion related impulsivity is more complex than an isolated cognitive process {gone awry (Johnson e* al, 2020), Itwil be important for researchers to move beyond these isolated, small effect sizes co integrate data from multiple cognitive domains using multivariate statistical methods. In this case, we would recommend that researchers use the DDT (unam Diguous and certain), RGF (unambiguous and uncertain), and IGT (ambiguous and uncertain) to cover the risky decision-making domain. Second, when studying emotion-related impulsivity in the ab, we recommend that future researchers carefully select and integrate effec tive arousal manipulations. The current study demonstrates that mon. ‘etary incentives and common postive mood induetions do not appear to licit riskier decisions among people with higher emotion-related impulsivity. However, the few studies using stress inductions, imag- inal sex, and pharmacology provide hints that stronger manipulations of physiological arousal may elicit stronger associations between risky ‘decision-making and emotion-elated impulsivity. Given the growing Dody of laboratory studies suggesting that symptoms are evoked after ‘mood inductions for those with this form of imprlsvity, and the large body of null findings with “cold” cognitive tasks (89°99 el, 201°, researchers in this field should effectively manipulate and measure physiological arousal Third, our analyses indicate that the correlation between risky decision-making and emotion-elated impulsivity generalizes across ‘demographic domains; however, gaps still remain with regard to racial ‘identity and lifespan development. The field will benefit from future studies that recruit racially diverse samples. It will also benefit from increasing the developmental range of inquiry in children, adolescents, and older adults. Its also likely that adversity, oppression, and add ‘onal forms of marginalization (eg, sexual and gender minority and disability status) impact processes relevant for both emotion-elated impulsivity (eg, Carver ct al, 2011) and decision-making (eg, Duly, Melaughn, ane Green, 2018) and should be evaluated as po tential moderator Fourth, more research is needed to determine whether increased risky decisionmaking contributes to poor psychosocial outcomes ‘experienced by individuals with high emotion-telated impulsivity. We ‘do not want co label the tendency toward risky decision-making 85 maladaptive per se, especially given the modest size ofthe effect found here. Nonetheless, given that some tasks included inthis meta-regression have been linked to real-world reckless driving (Sro%n et al, 20102, problem gambling (lcs ta, 2012), risky sexual choices Dere‘inks otal, 2014), and Binge drinking (Xis0 et 3, 2008), over time sky decision-making may be connected to the distress and functional impairment experienced by many with high emotion-rolated impulsivity. As the current study fills an important gap inthe literature on the ‘cognitive correlates of emotion-rlated impulsivity, so does it highlight the complexity ofthis critical puzzle. The symptomatle manifestations of cal rycen Review 10 (2028) 102232 ‘emotion elated impulsivity ae state-dependent, and most research on risky decision-making has failed to consider ths issue. Emotion-elated {Impulsivity also appears tobe linked to multiple cognitive domains, with "sky decision-making being just one domain implicated. It may even be the case that its cognitive correlates manifest in an equifinal manner, Therefore, it is unlikely that emotion-elated impulsivity will be captured by a single index or task even under optimal psychometric and physiological conditions. The findings in this systematic review and reta-regresson lead us to believe that risky decision-making tasks will, be valuable components of future, multivariate research that will ulti= rately unlock the neurocognitive code of emoton-related impulsivity, and lead us to more targeted, effective, and trensdiagnostic clinical Incerventions. Role of funding sources Sheri L. Jonson is the recipient of NIE grant ROIME110$77. Funding agencies had no role in the study design, analysis, or inter- pretation of the data, writing the manuscript, or the decision to submit the paper for publication. Contributors SJ, DML, and HK. designed the study and wrote the pre- registration. D.MLL. conducted literature searches and curated the ati- cles, ME, Su, and JP. reviewed the articles for inclusion and extracted the relevant effect sizes, M.E. and S.J. corresponded with authors of included studies to request missing data. MLE, S.J, and H.K. decided on the statistical methods. ME. wrote the analysis code and completed formal analysis. ME, SJ, and JP. wrote the first draft of the manu seript. All authors contributed to and have approved the final manuscript Declaration of Competing Interest The authors declare that they have no conflicts of interest Data availability Links to the prezegistration, data, code, and analysis output, wich have been posted on the Open Science Framework, are included at the beginning ofthe Method section within the manuscript Acknowledgements We would like to thank all our colleagues that contributed to the esearch that was synthesized inthis study, especially those that took the sime to send us data or to re-analyze their data so that it could be Included Appendix A. Supplementary data ‘Supplementary data to this article can be found online at hisp=://Aok. 0rg/10.1016/;.

    You might also like